Table 2.
Variable | Label | Odds Ratio (95% CI) | ||
---|---|---|---|---|
Model 1 | Model 2 | Model 3 | ||
3M ROM | Major vs. Mild or Moderate | 3.94*** (2.85–5.44) | 2.74** (1.96–3.82) | 2.74*** (1.97–3.82) |
Extreme vs. Mild or Moderate | 12.88*** (9.19–18.07) | 6.83*** (4.78–9.75) | 6.87*** (4.80–9.82) | |
Gender | Male vs. Female | 1.17 (0.92–1.48) | 1.08 (0.85–1.37) | 1.06 (0.83–1.35) |
Age (years) | 60–69 vs. < 60 | 1.17 (0.75–1.82) | 1.12 (0.72–1.75) | 1.03 (0.68–1.66) |
70–79 vs. <60 | 1.37 (0.92–2.05) | 1.46† (0.97–2.19) | 1.35 (0.90–2.04) | |
≥80 vs. <60 | 2.06*** (1.44–2.95) | 2.28*** (1.59–3.27) | 2.02*** (1.38–2.94) | |
Lab severity | Moderate vs. Mild | 1.75* (1.06–2.88) | 1.77* (1.07–2.90) | |
Severe vs. Mild | 5.52*** (3.48–8.75) | 5.65*** (3.56–8.97) | ||
Race/ethnicity | Chinese vs. Caucasian | 0.94 (0.57–1.56) | ||
Filipino vs. Caucasian | 0.69* (0.48–1.00) | |||
Hawaiian vs. Caucasian | 0.68* (0.47–1.00) | |||
Japanese vs. Caucasian | 0.83 (0.60–1.17) | |||
Other Pacific Islander vs. Caucasian | 0.35* (0.15–0.83) | |||
Other vs. Caucasian | 0.55* (0.32–0.94) | |||
Summary Statistics | ||||
AUC: mean ± SE (95% CI) | 0.775 ± 0.012 (0.751–0.799) | 0.815 ± 0.011 (0.794–0.836) | 0.819 ± 0.011 (0.798–0.840) | |
(-2)*log likelihood | 2,160.08 | 2,055.17 | 2,041.85 | |
Model d.f. | 6 | 8 | 14 | |
Δχ2 | 306.30 | 101.90 | 13.32 | |
p-value | <0.001 | <0.001 | 0.038 | |
10-fold Cross-Validation | ||||
AUCcv: mean ± SD | 0.773 ± 0.026 | 0.810 ± 0.021 | 0.810 ± 0.021 |
Notes: p < 0.10
p < 0.05
p < 0.01
p < 0.001.
Model 1 = 3M Risk of Mortality (ROM) + Gender + Age.
Model 2 = 3M ROM + Gender + Age + Lab Severity.
Model 3 = 3M ROM + Gender + Age + Lab Severity + Race/Ethnicity.
3M ROM = risk of mortality assigned by 3M based on administrative/claims data only.
Lab Severity = total number of abnormalities from admission lab at last hospitalization (the number of lab abnormalities was categorized by tertiles with the minimum and maximum values: Mild = 0–5, Moderate = 6–8, and Severe = 9–20).
AUC = area under the curve of predicted values or c-statistic.
Δχ2 = difference in (-2)*log likelihood between models.
Model d.f. = the number of model parameters.
p-value = p-value of Chi-squared (or log likelihood) test comparing two models.
Model 1 was compared with the intercept-only model.
AUCcv = area under the curve of predicted values from 10 validation sets.